The Hidden AI Tax: Your Electric Bill Could Become the Next Battleground of the Tech Boom
Artificial intelligence is most often talked about in terms of Nvidia chips, their trillion-dollar valuations and their huge data centres. The less exciting consequence could one day end up in the mailboxes of Americans in the form of a higher electric bill.
Artificial intelligence has now given rise to a rather odd kind of inflation. Since the technology industry requires huge quantities of electricity, the utilities will need billions of dollars worth of new infrastructure in order to meet this demand, and regulators are now being asked a politically difficult question: who should bear the cost of all the things needed to keep the AI boom going?
While American households are going through that discussion, electricity prices have already risen. According to the latest Consumer Price Index figures from the U.S. Bureau of Labor Statistics, electricity prices were 3.8% higher in August than they had been a year before. Data separately provided by the U.S. Energy Information Administration indicate that the average residential price of electricity in the first six months of 2026 was 18.16 cents per kilowatt-hour, 7.4% higher than it had been during the corresponding period in 2025.
It cannot be merely attributed to artificial intelligence. Factors such as fuel prices, the weather, investments in transmission, maintenance, wildfire protection and regional regulations all have an effect on electricity bills. However, AI is introducing something fundamentally different to the situation: huge new customers who can use electricity on a scale usually linked with cities and heavy industry. The point therefore facing investors and households is no longer whether artificial intelligence needs huge quantities of electricity since it clearly does; the issue is whether the enormous cost of providing that power stays with the technology companies which are demanding it or instead begins to appear on the electricity bills of everyone else.
AI Is Quietly Moving Wall Street’s Buyback Machine Downstream →
The figures relating to the expansion of electricity use are now hard to ignore. According to the newly updated U.S. data centre energy report from the Lawrence Berkeley National Laboratory, data centres could use 649 terawatt-hours of electricity by 2030, which would amount to about 11.8% of all electricity consumed in the United States. The laboratory’s different scenarios vary the figure between 9.5% and 15.3% of total U.S. electricity use, depending on how fast computing equipment is installed and how heavily the AI servers operate.
It is a significant change for an industry which, up to now, has accounted for only a relatively small share of the nation’s total electricity demand. The U.S. Department of Energy had already cautions in 2024 that domestic use of electricity by data centres could double or triple by 2028. More recent forecasts now indicate that the pressure could carry on long after that date.
The shortage is already being reflected in financial markets. As MarketWatch points out, Morgan Stanley has estimated that the United States might experience a 33-gigawatt power shortfall through 2028 since demand for computing infrastructure is outpacing the available electricity supply. This shortfall is causing power generation, turbines, grid equipment, transmission capacity and companies with access to secured electricity to become another area of focus in the AI investment sector.
On Wall Street, this could lead to potential winners, but the situation for households is more complicated. Somebody has to finance the generating capacity, the substations, the transmission lines and all the other upgrades needed to meet the huge new demand. Generally, utilities recover the costs of approved infrastructure through customer rates over time. However, if the regulators decide that the infrastructure is beneficial to the wider grid, some of those expenses might be shared among a large number of customers rather than being charged solely to one data centre. An AI infrastructure boom can then turn into an inflation story.
Why Everything Feels Expensive Again: Inflation, Gas Prices and the Fed’s Next Move →
Federal regulators have now admitted outright that there is a risk involved. In June the Federal Energy Regulatory Commission instructed all of the six regional grid operators falling within its authority to either justify or amend the rules concerning how data centres and other exceptionally large users of electricity connect to the grid. The FERC specifically stated that its move was intended to speed up connections while at the same time protecting standard ratepayers.
FERC Commissioner David Rosner was even more clear about the aims of the regulators. When speaking about the new cost-recovery requirements for large power users, he stated that “Cost Recovery Agreements prevent those costs from being shifted to residential customers.” He went on to say that the safeguards are especially important in the case of a utility constructing expensive infrastructure for a proposed data centre which is delayed or never comes about.
The fact shown highlights the underlying economic problem in the competition over AI. A technology company may announce the construction of a campus worth many billions of dollars long before the facility is fully operational. As a result, the utilities might be forced to start upgrading the local power grid long before time. In the event that the anticipated demand does not materialize, someone will still have to pay for infrastructure which has already been built.
Meanwhile, the utility companies are making profits which are the highest they have ever seen. According to the Financial Times, U.S. utility profits hit a record $63.1 billion in 2025, although regulators and elected officials in many states are becoming more critical of the returns being asked for and the proposed rate increases. The utilities say that since it is necessary to modernise an outdated grid and at the same time meet a huge amount of new demand, they need access to capital and to have operators who are financially sound. Critics, on the other hand, are becoming more doubtful about how much of this responsibility should be passed on to residential customers who are captive.
The tension might end up being one of the most neglected economic consequences of the AI boom, even if consumers never buy an Nvidia GPU, never subscribe to an advanced AI service or never invest in a technology stock, since the physical infrastructure that supports these businesses can still have an impact on the energy system in the surrounding area.
Energy Affordability Package: California’s $18 Billion Push to Cut Bills and Tame Wildfires →
A important counterpoint is that expanding electricity infrastructure is not necessarily adverse to households. Large data centres can increase a utility’s customer base, generate long-term electricity demand and possibly distribute the existing fixed costs among a larger number of users. Moreover, technology companies are becoming more likely to enter into power-purchase agreements, fund power generation projects and adopt on-site electricity sources in order to reduce their reliance on the current grid capacity.
There is more to the need for additional power generation and transmission than just the issue of artificial intelligence. The expansion of manufacturing, the growth of electric vehicles, the process of electrification and population increase are all causing electricity demand to change. According to the Energy Information Administration electricity demand has risen steadily since 2020 following more than ten years of very slow growth, current data centres being one important driver of this rise.
Greater investment could in the end result in a bigger, more efficient and more resilient electric system, and if the regulators set the rates properly, homes could gain from infrastructure that is in part financed by some of the richest corporations in the world. That distinction is important because the fact that there is an AI boom does not in itself mean that AI will cause each and every household’s electricity bill to rise. What it does mean is that regulators have to decide, on a project-by-project and market-by-market basis, which of the costs were incurred in order to serve a very large number of new customers and which costs are properly attributable to the wider electricity system.
The impact of that battle is also significant for investors since utilities may benefit from years of capital expenditure and from rising electricity demand, but a regulatory response concerning affordability might reduce returns, stop certain costs from being approved or force more of the infrastructure expenses onto data-centre operators themselves. Thus the same AI surge which is causing unprecedented demand for electricity could turn out to be a source of both higher earnings and regulatory risk.
It follows that the most important story concerning AI and inflation might have only a minor connection with the price of a semiconductor. It could end up being just one line on the monthly utility bill. The United States aims to develop sufficient computing capacity in order to take the lead in the next generation of technology while at the same time setting up the physical energy system needed to power it. According to Lawrence Berkeley’s projections, data centres could end up using about one out of every nine kilowatt-hours generated in the country by 2030 in the reference scenario. The FERC is already altering the rules concerning these very large new power users since the issue of how costs should be allocated can no longer be put off.
For investors there could be a completely new range of opportunities in the field of AI infrastructure due to the electricity shortage. For everyone else, there is a simpler question worth watching: When the bill for the AI revolution eventually comes, who is actually going to have to pay it?